Low- versus High-Dose and Early versus Late Parenteral Amino-Acid Administration in Very-Low-Birth-Weight Infants: A Systematic Review and Meta-Analysis

During the first weeks of life, premature neonates are at highest risk of developing nutritional deficits. These nutritional deficits may reach a daily shortage in energy and proteins in comparison to the recommended dietary intake [1]. Providing proteins or amino acids to preterm infants during the early postnatal period is critical for adequate growth and neurodevelopment [2,3]. Reduction in malnutrition results in higher postnatal growth rates and is associated with favorable long-term neurodevelopmental outcome [4,5,6,7,8]. Supplementation of amino acids has been shown to improve protein balance by increasing protein synthesis, improving the antioxidant defense system [9], and potentially preventing a catabolic state and neonatal growth retardation [10,11]. However, no consensus has yet been reached about what dose is appropriate or when to initiate supplementation.

Recent studies showed that a protein dose of 3.5 or even 4.0 g/kg/day is well tolerated [4,12,13]. This statement is in contrast with suggestions from earlier studies [14,15] that raised concerns about the incidence of metabolic acidosis and hyperammonemia in preterm infants receiving high doses of amino acids through parenteral nutrition. In addition, there is recent evidence suggesting that high doses (3.5 g/kg/day) of amino-acid supplementation do not improve neonatal growth compared to low doses (2.5 g/kg/day) [13,16]. As a result, there is no consensus about the optimal dose of parenteral amino-acid administration, there is a concern regarding side effects when higher doses are submitted, and, finally, it is uncertain whether neonatal growth improves at all. There is little assessment of long-term efficacy and safety, which makes reliable recommendations difficult.

Consequently, the first objective of this systematic review and meta-analysis is to identify the most suitable dose of parenteral amino-acid administration to very-low-birth-weight (VLBW; ≤1,500 g) infants within the first days of life based on short-term safety and efficacy and, when possible, to include long-term data, too. Anthropometric data and side effects associated with low- and high-dose amino acid supplementation were investigated to determine the efficacy and safety of this intervention.

In addition, there is as yet no determination of the optimal timing for initiating amino-acid administration to premature infants. Delaying amino acid intake in a preterm infant could potentially result in a catabolic state of nutrition and can lead to early postnatal growth failure. The time frame for regaining birth weight (BW) could be longer and the catch-up requirements greater [17]. In clinical practice, the start of amino-acid administration varies among institutions, and it is, therefore, important to critically review the evidence and assess the benefits and risks of early versus late amino-acid administration [18,19,20]. Subsequently, the second objective of this systematic review is to determine if early (starting within 24 h after birth) compared to late (starting later than 24 h after birth) supplementation of amino acids has an effect on VLBW infants' growth and to define the benefits and risks of this intervention.

For this systematic review and meta-analysis, the requirements of the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement were followed [21].

Searches in PubMed (http://ncbi.nlm.nih.gov/pubmed), EMBASE (http://www.embase.com), and Cochrane Central Register of Controlled Trials (CENTRAL, The Cochrane Library, http://www.thecochranelibrary.com, latest issue) through September 21, 2016, were done by using the key terms (words in the title or abstract of the study) “protein,” “peptide,” “amino acid,” “parenteral,” “intravenous,” and “infusion,” and the population-related terms “very low birth weight,” “preterm,” “premature”, plus “infant,” and “neonate.”

A manual search of reference lists of all relevant studies was performed by E.K.S.M.L. and checked by M.W. The searches were limited to human studies, and no language restriction was applied. However, studies written in languages other than English were excluded later in the process. The citations with abstracts were uploaded into a reference database (Reference Manager, version 11.0) and checked for duplicates.

E.K.S.M.L. and M.W. independently selected the studies. Studies were included if they met all of the inclusion criteria: a randomized controlled trial (RCT) design; a study group of preterm infants weighing <1,500 g who were admitted to a neonatal intensive care unit, needed parenteral nutrition, and received any type of parenteral amino-acid solution within the first days of life; and weight gain included as an outcome measure. No restriction on the dose of amino acids was applied. Cohort studies, case series or reports, and trials including only infants with congenital abnormalities were excluded.

Both reviewers read the selected articles. E.K.S.M.L. extracted, assessed, and coded all data for each study by using a form designed specifically for this review. For each study, E.K.S.M.L. entered final data into RevMan (version 5.2, 2012; The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Denmark). M.W. checked each stage of the process.

The following study data were extracted from the studies included:

1. Title, first author, journal, and year of publication

2. Study design and study participants' characteristics (number of participants, gestational age [GA], BW), and inclusion and exclusion criteria per study

3. Type of intervention and control treatment (duration, start of amino-acid administration, composition and initial and final dose of amino acids administered, and co-interventions)

4. Short-term outcome measures:

a. Primary efficacy outcomes: anthropometric data (weight gain, head circumference [HC] gain, linear growth, and lower-limb length) and time to regain BW

b. Secondary safety outcomes: death, duration of respiratory support and supplemental oxygen, necrotizing enterocolitis (NEC), sepsis, intraventricular hemorrhage (IVH), metabolic acidosis, hypo-/hyperglycemia, postnatal steroid use, blood urea nitrogen (BUN), hypoalbuminemia, hypocalcemia, hypophosphatemia, nitrogen balance, and creatinine and ammonia levels

5. Long-term outcomes: anthropometrics; neurodevelopmental outcome at ≥12 months of corrected age

The level of evidence of each article was established following the Oxford Center for Evidence-Based Medicine Level of Evidence scale. RCT quality was assessed by E.K.S.M.L. and M.W. using the Jadad criteria (0- to 5-point rating scale, with 5 as maximum score) [22].

Review Manager software was used to perform analyses (RevMan, version 5.2, 2012; The Nordic Cochrane Centre). A 2-sided value of p < 0.05 was considered significant.

The first comparison of this study was a higher-dose amino-acid administration (>3.0 g/kg/day) versus a lower dose (≤3.0 g/kg/day). A second comparison was performed to compare early (≤24 h) to late (>24 h) initiation of amino acids in VLBW infants. For both comparisons, all outcomes were assessed. To analyze treatment effect and calculate a pooled mean of outcomes reported in ≥2 studies, the Mantel-Haenszel method was used for categorical outcomes, and the inverse variance method was used for continuous outcomes.

For all outcome measures, we assessed the statistical heterogeneity of data from the included studies by calculating the I² statistic (I² >50% was considered heterogeneous).

In cases of poor study quality, a sensitivity analysis was performed by removing the particular study to examine whether this significantly changed the results.

A total of 3,211 titles and abstracts were screened in the databases, and 2,270 individual articles were identified. Six additional articles were found through a search of reference lists. After screening titles and abstracts, 2,119 articles were excluded. The remaining 157 full-text articles were assessed for eligibility (Fig. 1).

Sixteen of the 157 studies met the predefined inclusion criteria (Table 1). Eight studies compared low- versus high-dose amino-acid supplementation [13,16,23,24,25,26,27,28], 5 studies compared early versus late supplementation [12,29,30,31,32], and 3 studied both [16,33,34,35]. Three studies, from Blanco et al. [33,34,35], were based on the same study population and were, therefore, seen as one. Six studies included co-interventions in their feeding protocol: 2 compared early versus late amino-acid supplementation [12,32] and 4 studied low- versus high-dose supplementation [13,25,27,28]. All authors were contacted for additional nonpublished data on growth and secondary outcome measures to enhance the amount of consistently defined data for meta-analysis.

To detect publication bias, a funnel plot was constructed. However, there was an insufficient number of studies to permit proper evaluation of publication bias and to evaluate potential asymmetry of the funnel plot using Begg and Egger tests.

Characteristics of the included studies assessing the effect of low- (<3.0 g/kg/day) versus high-dose (>3.0 g/kg/day) parenteral amino-acid supplementation are shown in Table 1. The inclusion of preterm infants was based on GA in 3 studies, BW in 4 studies, and both BW and GA in 2 other studies.

The maximum doses in the intervention groups ranged from 3.5 to 4.0 g/kg/day. In 6 of 9 studies, the dose was slowly increased by 0.5-1.0 g/kg/day over several days, starting with doses between 1.0 and 3.0 g/kg/day on the first day of life (DOL). Morgan et al. [25], Vlaardingerbroek et al. [13], and Uthaya et al. [28] started with the maximum doses of 3.8, 3.6, and 3.6 g/kg/day on DOL 1, respectively.

Three studies scored 5 points on the Jadad assessment [22]. For the other studies, points were subtracted because the randomization method was not discussed, the study was not double blind, or no appropriate blinding method was described (Table 2). Dropouts and withdrawals were described in all but 1 study. Baseline characteristics and outcome measures of the studies are shown in Table 3. Blood levels of ammonia and creatinine, nitrogen balance, duration of oxygen support, and incidences of metabolic acidosis and hypoglycemia are not shown because none or only 1 of the included studies discussed these outcomes. Overall, 1,149 infants were included in the studies: 573 in the intervention groups and 576 in the control groups.

Primary Outcome Measures

In all studies, weight gain was described as an outcome measure. Tan and Cooke [27] did not find a difference in weekly weight gain rates between the 2 groups in the first 7 weeks of life but did not specify exact growth rates. Burattini et al. [24], Clark et al. [16], Vlaardingerbroek et al. [13], Blanco et al. [33,34,35]. and Bellagamba et al. [23] reported actual rates in gram per kilogram per day. A meta-analysis including these 5 studies showed no differences between the intervention and control groups (Fig. 2a). However, a lack of consistency in the timeline of this outcome could have influenced the results. Vlaardingerbroek et al. [13], Clark et al. [16], and Blanco et al [33,34,35] reported mean weight gain for the first 28 days, Burattini et al. [24] compared mean weight gain at 36 weeks postmenstrual age (PMA), and Bellagamba et al. [23] reported mean weight gain rates from birth and regain of BW to 1,800 g (54 days in both groups). Scattolin et al. [26] also reported growth rates, but only for the 2nd and 3rd week of life separately and was, therefore, not included in the meta-analysis. They described a higher growth rate in the intervention group that became significant in the 3rd week of life (18.76 [SD 6.83] vs. 14.70 [SD 8.99] g/kg/day, p < 0.01).

HC growth rates were measured in 3 studies [13,16,27]. Meta-analysis was not possible due to a lack of consistency in the timeline, but none of the individual studies found a significant difference in head growth rate between groups. Mean HC per group at 36 weeks PMA was described in 5 studies [23,24,25,26,27]. All 5 studies indicated no difference in HC at birth. One study (Morgan et al. [25]) described a significant difference in mean HC in favor of the intervention group (p = 0.007). However, a meta-analysis of the 5 studies showed no significant difference in mean HC at 36 weeks PMA between groups (p = 0.99, with I² = 39%; Fig. 2b). One study described HC at term age and reported on a smaller mean HC in the intervention group (mean difference: 0.8 cm, p = 0.02) [28]. Finally, Blanco et al. [33,34,35] described no significant difference in HC between groups at birth and discharge.

Mean linear growth rates (mm/day) using lower-limb-length gain, were measured in 2 studies [13,27]. In Tan and Cooke [27], lower-limb-length gain was higher in the intervention group in the first 2 weeks of life (0.28 vs. 0.20 mm/day, p = 0.05). Vlaardingerbroek et al. [13] did not find a difference between groups (mean 0.26 [0.16 SD] vs. 0.27 [0.13 SD] mm/day) in the first 28 days of life. No meta-analysis was possible due to a lack of consistency in the definition of the outcome measure. Mean length at 36 weeks PMA was described in 4 studies [23,24,26,27]. In 3 studies, no significant differences were found. Tan and Cooke [27] found a significant difference in mean length at 36 weeks PMA in favor of the intervention group. Meta-analysis showed no significant difference between the groups (p = 0.09, Fig. 2c).

The same 4 studies [23,24,26,27] reported the mean number of days to regain BW. Meta-analysis showed no significant difference between both groups (p = 0.20; Fig. 2d). The heterogeneity of these studies was shown by an I² of 66%.

Secondary Outcome Measures

All studies reported on death during hospital admission. In the individual studies, no significant differences in overall mortality were found. In total, 50 of 491 (10%) participants in the high-dose group died vs. 60 of 494 (12%) participants in the low-dose group. Meta-analysis did not show a significant difference in mortality before discharge or DOL 28 (p = 0.86; Fig. 2e).

Serum BUN levels were reported in 7 studies. All studies found elevated levels in the intervention group in comparison to the control group. Blanco et al. [33,34,35] measured BUN levels at DOL 1 and 7 and found a significant difference on both days (mean differences of 1.4 and 6.4 mmol/L, respectively, p< 0.001). Clark et al. [16] reported a significant difference between groups measured on DOL 7 (p = 0.01). Bellagamba et al. [23] and Burattini et al. [24] found a significantly higher BUN level in the intervention groups throughout the first 5 and 10 days of life, respectively. Uthaya et al. [28] found that significantly more infants had BUN levels of >7 and >10 mmol/L in the intervention groups (p < 0.01). Vlaardingerbroek et al. [13] and Scattolin et al. [26] reported serum BUN levels on days 2, 4, and 6, and 7 and 21, respectively. A meta-analysis was performed for days 1-2 and 6-7 (Fig. 2f, g), which showed a significant effect on days 1-2 (p = 0.01) and days 6-7 (p = 0.03). Heterogeneity of the studies was shown by an I² of 78 and 90%, respectively (Fig. 2f, g). Nitrogen balance was addressed in 1 study [13], which reported a significantly higher positive balance on DOL 2 in the intervention group in comparison to the control group. This difference disappeared on days 4 and 6. Two studies reported on creatinine levels [16,26]; neither study found a significant difference between intervention and control groups. Because of a lack of consistency in outcome measures, it was not possible to perform a meta-analysis.

The definition of hyperglycemia differed between studies. Tan and Cooke [27] defined hyperglycemia by a blood glucose level >12 mmol/L that was treated with insulin. They found a significant difference between the number of infants that needed insulin treatment in the groups - 33 in the intervention group vs. 12 in the control group (p < 0.01) - while dextrose intake was higher in the intervention group. Burattini et al. [24] defined hyperglycemia as blood glucose levels >175 mg/dL (∼9.7 mmol/L) at 2 consecutive measurements. They also found a significant difference, but in favor of the intervention group, which had significantly fewer episodes of hyperglycemia (6 vs. 20) during hospital admission (p < 0.001). Glucose intake was marginally but significantly higher in the intervention group. Scattolin et al. [26], Vlaardingerbroek et al. [13], Uthaya et al. [28], and Blanco et al. [33,34,35] found no differences between groups. Blanco et al. [33,34,35] reported no actual data with this statement. Meta-analysis was possible for 2 studies with a similar glucose intake between groups [13,24] and showed an association between high-dose amino-acid supplementation and a lower incidence of hyperglycemia (p = 0.02, I² = 70%; Fig. 2h).

The number of infants needing oxygen supplementation at 36 weeks PMA was reported by 3 studies. Blanco et al. [33,34,35] found a significant difference in favor of the control group (p = 0.01). Burattini et al. [24] and Tan and Cooke [27] did not find a difference. Meta-analysis showed no overall effect (p = 0.5, with an I² of 73%; Fig. 2j). Four studies reported on duration of mechanical ventilation. Lack of consistency in outcome measures made meta-analysis impossible. Individual studies did not show a significant effect.

All studies reported on NEC incidence. Studies used different definitions, including need for surgical intervention, strong clinical suspicion leading to medical treatment, and Bell stage >II. In 4 studies, no definition was given. None of the individual studies reported on a significant difference, which was confirmed by meta-analysis including 966 infants (p = 0.80; Fig. 2k).

Additionally, all studies reported on sepsis incidence. Vlaardingerbroek et al. [13] defined sepsis according to criteria described by Stoll et al. [36]. Additionally, a blood culture positive for coagulase-negative staphylococci together with elevated C-reactive protein (>10 mg/L) was considered true sepsis. Scattolin et al. [26] defined sepsis as deterioration of clinical condition with a positive blood culture. The remaining studies defined sepsis as a positive blood, urine, or cerebrospinal fluidculture or did not report a definition. None of the studies found a significant difference between groups. Meta-analysis underlined this finding (p = 0.95; Fig. 2l).

Eight studies reported on IVH grade ≥III. No significant differences in overall incidence were found. This was supported by meta-analysis (p = 0.99; Fig. 2m). Neurological outcome at 2 years of age was studied by Blanco et al. [33,34,35], Burattini et al. [24], and Bellagamba et al. [23]. No significant differences were found between groups.

Postnatal steroid use was reported by 5 studies. Individual studies did not find a significant difference between groups. Meta-analysis supported this finding (p = 0.39, I² = 45%; Fig. 2n).

Data on metabolic acidosis was reported by Bellagamba et al. [23]. Metabolic acidosis was defined as a standard base excess <7.5 mmol/L in 2 consecutive gas analyses. Bicarbonate treatment was started when these conditions were met. Bicarbonate treatment during parenteral nutrition was required in more infants in the intervention group than in the control group (73 vs. 63; p = 0.038).

Data on hypoglycemia, hypoalbuminemia, hypocalcemia, hypophosphatemia, or ammonia levels were reported by only 1 study or none.

After removing the study with a Jadad score <3 [26], serum BUN levels on days 6-7 were no longer significantly different between groups (p = 0.10). All other results did not change.

Characteristics of the included studies assessing the effect of early (<24 h after birth) versus late (>24 h) introduction of parenteral amino acids are shown in Table 1. In 2 studies, the included infants had a BW of <1,500 g, in 1 study the infants had a BW of <1,000 g, and in the remaining studies inclusion was based on GA rather than BW.

Amino-acid supplementation was initiated on the first DOL in all intervention groups and in the control groups on DOL 2-4. Ibrahim et al. [12], Heimler et al. [29], and Wilson et al. [32] described unblinded RCT. Te Braake et al. [31] used a single-blinded approach, while the approach of Blanco et al. [33,34,35] was double blinded but the blinding method was not mentioned. In Saini et al. [30], blinding was not mentioned. Four of the 6 studies were randomized in an appropriate way according to the Jadad criteria [22]. Te Braake et al. [31] did not describe their randomization method, and Saini et al. [30] randomized sequentially and, therefore, inappropriately (Table 2).

Baseline characteristics and outcome measures of the studies are shown in Table 4. Creatinine, postnatal steroid use, hypoglycemia, and metabolic acidosis are not included in this table because the included studies did not report data on these outcomes. Overall, 405 infants were included in the studies: 203 in the intervention groups and 202 in the control groups.

Primary Outcome Measures

In 3 of 6 studies (Blanco et al. [33,34,35], Ibrahim et al. [12], and Saini et al. [30]), mean weight gain was given as outcome measure for growth. No statistical difference between groups was found by the individual studies. Due to a lack of consistency in the published data, no meta-analysis could be performed. Additional nonpublished data on growth became available for 1 study [33,34,35], but this was insufficient for further statistical analysis.

HC was measured by Heimler et al. [29] and Saini et al. [30]; neither found a significant difference between groups. Heimler et al. [29] described a mean HC increment in the first 2 weeks of 0.50 cm (0.50 SD) in the intervention group vs. 0.25 cm (0.27 SD) in the control group (p = 0.22). Saini et al. [30] found a mean HC increment of 0.40 cm (0.1 SD) in the first 10 days in both groups.

Number of days to regain BW was assessed by Te Braake et al. [31], Heimler et al. [29], and Wilson et al. [32]. The latter reported, as a single study, a significant difference between groups in favor of the intervention group (median 9 vs. 12 days, p < 0.001). Because of a lack of consistency in these measurements, meta-analysis was not possible.

Secondary Outcome Measures

Death during hospital stay was reported in 4 studies. Meta-analysis did not show a significant effect of early introduction of amino-acid supplementation on overall mortality (p = 0.63; Fig. 3a). Individual studies either did not find significant differences in overall mortality or did not report a p value.

Serum BUN levels were reported by Blanco et al. [33,34,35], Te Braake et al. [31], and Heimler et al. [29]. All found significantly elevated serum BUN levels in the intervention groups in comparison to the control groups. First, Te Braake et al. [31] measured serum BUN on DOL 2, 4, and 6 and found differences (p < 0.05) on all days. Second, Heimler et al. [29] collected blood samples at a mean age of 78 h and found a mean difference of 4.0 mmol/L in favor of the intervention group (p < 0.001). Third, Blanco et al. [33,34,35], who measured levels on DOL 1 and 7, found differences on both days (mean: 1.43 and 6.4 mmol/L, respectively; p < 0.001). Meta-analyses performed for days 1-2, 3-4, and 6-7 showed a significant effect on days 1-2 (p = 0.02) and 3-4 (p < 0.00001), but not on days 6-7 (p = 0.09; Fig. 3b-d).

Nitrogen balance was described by Heimler et al. [29], Ibrahim et al. [12], Saini et al. [30], and Te Braake et al. [31]. Premature infants who received early amino-acid supplementation had a positive nitrogen balance, whereas a negative nitrogen balance was recorded in infants who received supplementation late (>24 h). In the study by Heimler et al. [29], urine was collected on DOL 3, before the administration of amino acids to the control group. The study showed a positive mean nitrogen balance in the intervention group and a negative mean balance in the control group. In addition, Ibrahim et al. [12] reported a significant difference between groups on days 1, 3, and 5 (urine was collected in 29 infants [90.6%]) but not on day 7. The intervention group started with a mean nitrogen balance of 384.5 mg/kg/day on DOL 1, which remained stable until day 7. The control group started with a negative nitrogen balance (mean -203.4 mg/kg/day), but the level increased until day 7. On day 7 a 4th measurement showed no statistically significant difference (no p value available). Te Braake et al. [31] reported a significant difference in favor of the intervention group on day 2. Contrarily, on day 4, a significant difference was found in favor of the control group. Saini et al. [30] described a significant effect in favor of the intervention group on days 1-3 but not thereafter. Meta-analysis showed a significant difference between groups: p = 0.0003 (Fig. 3e). χ² statistics revealed substantial heterogeneity (I² = 100%).

Serum glucose levels were reported by all studies. Wilson et al. [32] described hyperglycemia as blood glucose levels >11 mmol/L with glycosuria >+++ or 55 mmol/L. The number of infants requiring insulin treatment was reported separately. Only the latter was significantly different between both groups in favor of the control group, of which 2% were treated with insulin in comparison to 33% of infants in the intervention group. Blanco et al. [33,34,35] defined hyperglycemia as blood glucose levels >150 mg/dL and found no difference between groups (p = 0.51). Ibrahim et al. [12], Te Braake et al. [31], and Heimler et al. [29] did not report the incidence of hyperglycemia but rather the mean serum glucose values of both groups. Te Braake et al. [31] found a significantly lower value in the intervention group on DOL 2. Meta-analysis was not possible due to inconsistency in definitions.

Neurodevelopmental outcome after 2 years was reported by Blanco et al. [35]. No significant difference was found between groups. The same holds for the duration of mechanical ventilation, duration of supplemental oxygen, incidence of NEC, sepsis, IVH, and levels of ammonia. Te Braake et al. stated that there was no difference in metabolic acidosis between groups without presenting actual data. None of the included studies reported on hypoglycemia, postnatal steroid use, creatinine levels, hypoalbuminemia, hypophosphatemia, or hypocalcemia.

When removing the studies with a Jadad score <3 [12,29,30], only 1 meta-analysis (on overall mortality) could be performed. The outcome of this analysis did not change.

In this systematic review, the efficacy and safety of both high- versus low-dose and early versus late parenteral amino-acid supplementation in VLBW infants was studied. The benefits of these interventions, as well as the number and diversity of adverse events, were reported. The results suggested that administering a high dose (>3.0 g/kg/day) or an early dose (≤24 h) of parenteral amino acids is well tolerated, but does not offer significant benefits on growth when compared to a lower dose or later initiation. On the other hand, it does not cause a higher incidence of adverse events either.

Unfortunately, for several outcome measures, a (meaningful) meta-analysis could not be performed due to considerable heterogeneity in reported outcome measures in the included RCT despite our attempt to include additional, nonpublished data by contacting the authors. Furthermore, Jadad quality ratings of the included studies varied, and removal of low-quality studies for sensitivity analysis resulted in even less obtainable meta-analyses.

One possible reason why higher doses of amino acids did not translate into improved growth outcomes might be that the solutions used were of suboptimal composition for the needs of preterm infants. Limited data were available on the parenteral amino-acid compositions used by the different RCT. In 2 studies [13,31], the source was Primene 10% (Baxter Healthcare Corporation), in 1 study Aminosyn PF (Abbott Laboratories, Chicago, IL, USA) [33,34,35], in 2 studies TrophAmine (Baxter Healthcare Corporation or McGaw Inc.) [24,29], and in another study both Aminosyn and TrophAmine [16]. Due to this variability, comparing the results across studies is difficult. Aminosyn, for example, only contains essential amino acids, while Primene and TrophAmine both contain essential and nonessential amino acids. None of the studies compared different compositions for superiority; therefore, no definitive statement could be made on which amino-acid composition should be used. All studies compared plasma concentrations of amino acids to previously reported levels in healthy term breastfed infants. For example, in Clark et al. [16], supplementation of amino acids resulted in excess levels of arginine, leucine/isoleucine, methionine, and ornithine but inadequate concentrations of tyrosine. Remarkably, the combined plasma concentrations of leucine and isoleucine were less than normal human fetal plasma concentrations [37]. This was documented for more (non)essential amino acids in other studies [10,38] and might suggest that the current solutions are not optimal for VLBW infants, since even high doses do not result in plasma amino-acid concentrations that are sufficient to promote growth. Nevertheless, Poindexter et al. [11] did document on significantly better growth outcomes at 36 weeks of PMA in VLBW infants who received greater early amino-acid supplementation (≥3 g/kg/day at ≤5 days of age) compared to lower late supply. These results were based on a secondary analysis (including 1,018 infants) of data from a RCT with a study design similar to that of Clark et al. [16]. Since the study maintained a higher total protein and amino-acid supply for a longer period, it has been proposed that both early and sustained higher rates of amino acids are necessary to achieve growth in VLBW infants [39].

On the other hand, the lack of growth improvement after higher-dose amino-acid supplementation has also been suggested to be due to the existence of a threshold above which the delivery of protein results in amino-acid oxidation without improvement in nitrogen retention or growth [13,28]. Based on fetal human data, this threshold is suggested to be ∼4 g/kg/day of protein [40]. Indeed, 7 studies on low- versus high-dose amino-acid supplementation found statistically significantly elevated serum BUN levels in the intervention groups in comparison to the control groups. In 1 study, the amino-acid intake of 23 infants (28%) was reduced due to elevated BUN levels in comparison to 14 infants (17%) in the control group (p = 0.093) [23]. In the study by Blanco et al. [33] in extremely low-birth-weight infants, 6 infants in the early and high-dose group (4 g/kg/day on day 3) were withdrawn from the study protocol due to elevated serum BUN levels and hyperammonemia. However, in this study, plasma amino acid concentrations were remarkably higher than those reported by others at the same dose of amino acids [10,16,34], which may reflect an inefficient metabolism in these infants. Also, the study by Blanco et al. [33] was the only one in which Aminosyn was used as the source of amino acids, which might provide different BUN and ammonia values from those using TrophAmine or Primene. Furthermore, the study has been criticized on its design [41], and the results have not been replicated by others. Higher BUN concentrations are to be expected when a higher dose of amino acids is administered immediately after birth, reflecting amino-acid oxidation and protein turnover and not toxicity. It remains debatable as to what constitutes a clinically significant state of uremia [42]. Very recently, van den Akker et al. [43] reported no association between early-life urea levels and neurocognitive outcome at 2 years of corrected age.

As a consequence of amino-acid oxidation after administration of high doses of amino acids, metabolic acidosis may occur [14,15]. Indeed, infants in the supplemented groups of 2 included RCT needed more sodium acetate or bicarbonate than infants in the control groups, suggesting lower pH values in the supplemented groups [23,33,34,35]. However, no adverse effects were described.

Meta-analysis of the data on early versus late amino-acid initiation also showed a significantly higher nitrogen balance in the early supplementation group (p = 0.003). Premature infants who received amino-acid supplementation early had positive nitrogen balances, and negative balances were recorded in infants who received supplementation later. The potential clinical significance of an early positive nitrogen balance is not known. Saini et al. [30] stated that it may be reasonable to assume an advantage, particularly during the critical transitional phase of extrauterine adaptation.

We found no significant differences in overall mortality, duration of mechanical ventilation or oxygen supplementation, incidence of NEC, sepsis, IVH, hyperglycemia, postnatal steroid use, or neurodevelopmental outcome for both research questions.

In 2013, a Cochrane review was published on early (<24 h after birth) versus late (>24 h) parenteral amino-acid supplementation in premature infants [20]. In addition, in a subgroup analysis, the authors compared high- (≥2.0 g/kg/day) versus low-dose (<2.0 g/kg/day) amino-acid supplementation. Consistent with our results, the authors found no short-term difference in length and HC between groups for both research questions. No statements were made regarding our other primary outcomes. In our review, 10 additional studies were included, of which 6 were recent studies. Therefore, we were able to make additional and more extensive statements on the outcome measures. Nevertheless, due to the great variability in study designs and outcome measurements as well as the general low quality of the available studies, it remained difficult to perform true meta-analyses. Furthermore, although differences in enteral diet, lipid intake, and energy dosing between studies may be confounding factors, no subgroup analyses could be done due to insufficient data or heterogeneity.

In summary, from available data, it can be concluded that a combination of early and high-dose supplementation of amino acids in VLBW infants is safe and well tolerated but yields no benefits on anthropometric outcomes when compared to later or lower-dose supplementation. It leads more rapidly to a higher nitrogen balance and results in higher serum BUN levels without detrimental effects. Although exact beneficial effects of these findings are not known, it is suggested that a more positive nitrogen balance has advantages in the transitional phase of extrauterine adaptation [30]. High serum BUN levels are presumably a reflection of a higher amino-acid oxidation rate and not a sign of intolerance [31].

This systematic review highlights the considerable variability in study design, interventions, and outcomes of available RCT, whereby meta-analysis of the results was often not possible, and thus no firm recommendations could be made.

Further research including larger numbers of infants is needed to study the effects of early and high-dose amino acids on both short-term and long-term anthropometric outcomes more consistently and extensively. These studies should also take the influence of parenteral lipids and dextrose, as well as enteral nutrition, into account. Finally, future research is warranted to determine the optimal amino-acid composition of parenteral nutrition.

The authors declare no conflicts of interest.

No funding was received for this project.

E.K.S.M.L. and M.W. share first authorship.